Monday, 5 April 2010

Base rates?

Kiwiblog nicely punches holes in ALAC's latest push for reduced drink driving blood alcohol limits.

I fear, though, that the stats are a bit of a nonsense without some measure of base rate. Let's take ALAC's numbers for the time being and ignore all of Farrar's very legitimate concerns. They say adult drivers with BAC between 0.05 and 0.08 are responsible for 7 of some 366 total road fatalities and some injuries. But we don't know what proportion of the overall driving population drives with blood alcohol between 0.05 and 0.08. If it's about two percent, then 7 of 366 doesn't seem all that out of line. If it's about half a percent, then 7 of 366 might be high, but we'd also have to adjust for time of day. All it would take would be to get aggregate data from the random police breath checks. If ten percent of the drivers on the road between midnight and 4 am have blood alcohol in the 0.05 to 0.08 range and if they're responsible for about ten percent of the accidents in that time, then we might be a bit overoptimistic if we figure we could cut down the road toll substantially by reducing the drink driving limit to 0.05: stone sober drivers also have accidents.

I still figure a reasonable move would be to combine any reduction in the blood alcohol limit with an opt-out procedure. Get a random sample of drivers and put them, sober, in a driving simulator. Tell them that the high scores get reduced licence fees so they take it seriously. Get an average score and the dispersion. Then, let anybody try for a drinking endorsement on the license. Anybody who sits the test while at 0.08 or higher and posts a score higher than a standard deviation below the sober mean gets to drive at that level of inebriation - a special endorsement on the licence. If that isn't good enough, why are we letting the bottom third or so of the distribution drive at all?

Previous posts:


  1. You are such a genius. Too bad the moralistic prigs in charge of the drunk driving programs real aim is to reducing drinking, not accidents.

  2. @dlr: reducing the rate to 0.05 would reduce both drinking and accidents; it just seems unlikely that it would prevent enough accidents to make it worth the cost in terms of consumption forgone.